Research On Image Denoising Algorithm Based On Multiscale Geometric Analysis

With the solid theoretical foundation and rigorous mathematical system, theWavelet transform has become a new field of the rapid development of the currentengineering application, which is considered to be the effective signal analysis toolafter the Fourier transform. However, with the development of the application of theWavelet transform, flaws and shortcomings have been exposed gradually in imageprocessing applications. The theory of multiscale geometric analysis provides thepossibility to make up for the inadequacy of the wavelet transform and provides a newmethod for sparse representation of high dimension signal. Multiscale geometricanalysis theory is in its infancy, however, which set off a revolution once again insignal analysis field after the Wavelet transform. Because of the superior nonlinearapproximation properties wavelet become an important method for image sparserepresentation. Image sparse representation is particularly important in the storage andtransmission. The proposal of multiscale geometric analysis theory provides anothereffective method for image sparse representation.This paper makes a systematic and thorough study of multiscale geometricanalysis theory with the image denoising in the background. On this basis, we studyWavelet transform and Contourlet transform and its application in image denoising.The main contents and research of the thesis are as follows:Firstly, the Wavelet transform and the multiscale geometric analysis theory areintroduced briefly. Multiscale geometric analysis tools are described and alsosummarized the advantages and defects of each tool, which provides theoreticalsupport and reference for the subsequent new algorithm.Secondly, standard procedure of image denoising algorithm based on Contourlettransform threshold is introduced. The selection of threshold is discussed andexpounded deeply. On this basis, an improved threshold image denoising algorithmbased on Contourlet transform is proposed, and its feasibility and effectiveness is verified by experiments.Finally, the applications of anisotropic diffusion in image processing areintroduced in detail. On this basis, an image denoising algorithm by integratinganisotropic diffusion with Contourlet transform was proposed, and its feasibility andeffectiveness is verified by experiments.